CN114660180A - Sound emission and 1D CNNs-based light-weight health monitoring method and system for medium and small bridges - Google Patents
Sound emission and 1D CNNs-based light-weight health monitoring method and system for medium and small bridges Download PDFInfo
- Publication number
- CN114660180A CN114660180A CN202210172090.7A CN202210172090A CN114660180A CN 114660180 A CN114660180 A CN 114660180A CN 202210172090 A CN202210172090 A CN 202210172090A CN 114660180 A CN114660180 A CN 114660180A
- Authority
- CN
- China
- Prior art keywords
- acoustic emission
- cnns
- damage
- model
- medium
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000013527 convolutional neural network Methods 0.000 title claims abstract description 55
- 238000012544 monitoring process Methods 0.000 title claims abstract description 37
- 238000000034 method Methods 0.000 title claims abstract description 33
- 230000036541 health Effects 0.000 title claims abstract description 30
- 230000006378 damage Effects 0.000 claims abstract description 53
- 238000012549 training Methods 0.000 claims abstract description 24
- 238000000354 decomposition reaction Methods 0.000 claims abstract description 6
- 238000002474 experimental method Methods 0.000 claims abstract description 4
- 238000011161 development Methods 0.000 claims description 26
- 238000012360 testing method Methods 0.000 claims description 21
- 230000006870 function Effects 0.000 claims description 20
- 230000008569 process Effects 0.000 claims description 10
- 238000011176 pooling Methods 0.000 claims description 9
- 230000000694 effects Effects 0.000 claims description 7
- 238000013528 artificial neural network Methods 0.000 claims description 6
- 210000002569 neuron Anatomy 0.000 claims description 5
- 230000004913 activation Effects 0.000 claims description 4
- 238000011156 evaluation Methods 0.000 claims description 4
- 239000011159 matrix material Substances 0.000 claims description 4
- 238000003062 neural network model Methods 0.000 claims description 4
- 125000004122 cyclic group Chemical group 0.000 claims description 3
- 238000005457 optimization Methods 0.000 claims description 3
- 238000003672 processing method Methods 0.000 claims description 3
- 230000002787 reinforcement Effects 0.000 claims 1
- 238000005516 engineering process Methods 0.000 abstract description 10
- 238000004364 calculation method Methods 0.000 abstract 1
- 238000001514 detection method Methods 0.000 description 7
- 238000013135 deep learning Methods 0.000 description 5
- 238000013136 deep learning model Methods 0.000 description 4
- 238000012545 processing Methods 0.000 description 4
- 230000003044 adaptive effect Effects 0.000 description 3
- 230000008859 change Effects 0.000 description 3
- 238000005070 sampling Methods 0.000 description 3
- 229910000831 Steel Inorganic materials 0.000 description 2
- 238000004458 analytical method Methods 0.000 description 2
- 238000013178 mathematical model Methods 0.000 description 2
- 210000000056 organ Anatomy 0.000 description 2
- 239000010959 steel Substances 0.000 description 2
- 230000007547 defect Effects 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 238000004880 explosion Methods 0.000 description 1
- 238000000605 extraction Methods 0.000 description 1
- 238000007689 inspection Methods 0.000 description 1
- 230000000306 recurrent effect Effects 0.000 description 1
- 230000009467 reduction Effects 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
- G01N29/00—Investigating or analysing materials by the use of ultrasonic, sonic or infrasonic waves; Visualisation of the interior of objects by transmitting ultrasonic or sonic waves through the object
- G01N29/14—Investigating or analysing materials by the use of ultrasonic, sonic or infrasonic waves; Visualisation of the interior of objects by transmitting ultrasonic or sonic waves through the object using acoustic emission techniques
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
- G01N29/00—Investigating or analysing materials by the use of ultrasonic, sonic or infrasonic waves; Visualisation of the interior of objects by transmitting ultrasonic or sonic waves through the object
- G01N29/44—Processing the detected response signal, e.g. electronic circuits specially adapted therefor
- G01N29/4409—Processing the detected response signal, e.g. electronic circuits specially adapted therefor by comparison
- G01N29/4418—Processing the detected response signal, e.g. electronic circuits specially adapted therefor by comparison with a model, e.g. best-fit, regression analysis
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
- G01N29/00—Investigating or analysing materials by the use of ultrasonic, sonic or infrasonic waves; Visualisation of the interior of objects by transmitting ultrasonic or sonic waves through the object
- G01N29/44—Processing the detected response signal, e.g. electronic circuits specially adapted therefor
- G01N29/4481—Neural networks
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
- G01N2291/00—Indexing codes associated with group G01N29/00
- G01N2291/02—Indexing codes associated with the analysed material
- G01N2291/023—Solids
Landscapes
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Immunology (AREA)
- Life Sciences & Earth Sciences (AREA)
- Chemical & Material Sciences (AREA)
- Analytical Chemistry (AREA)
- Biochemistry (AREA)
- Health & Medical Sciences (AREA)
- Engineering & Computer Science (AREA)
- General Health & Medical Sciences (AREA)
- Pathology (AREA)
- Signal Processing (AREA)
- Acoustics & Sound (AREA)
- Artificial Intelligence (AREA)
- Evolutionary Computation (AREA)
- Testing Of Devices, Machine Parts, Or Other Structures Thereof (AREA)
Abstract
The invention discloses a method and a system for monitoring light-weight health of a medium-small bridge based on acoustic emission and a one-dimensional convolutional neural network (1D CNNs), wherein the method comprises the following steps: arranging an acoustic emission sensor at a key vulnerable part of a medium-small bridge to acquire real-time data; carrying out empirical mode decomposition on the acquired acoustic emission data to achieve the purpose of denoising; and carrying out a reduced scale model experiment on the small and medium-sized bridges, and training the 1D CNNs model by using the acoustic emission signals of the reduced scale model experiment. And inputting signals acquired on the real bridge into the 1D CNNs model to monitor the damage stage of the structure in real time, and performing grading early warning. The invention provides a real-time lightweight health monitoring method and system for small and medium bridges by utilizing the advantages that the acoustic emission technology can monitor structural damage in real time and 1D CNNs can simplify the calculation complexity, and provides a basis for safety decisions of bridge management departments.
Description
Technical Field
The invention relates to the technical field of structural health monitoring, in particular to a monitoring method and a monitoring system combining acoustic emission and deep learning.
Background
The medium and small bridges are widely applied in daily life, and are damaged due to sudden or accumulated damages of different degrees when being exposed to outdoor environment in daily life, so that scientific and convenient safety assessment needs to be carried out on the medium and small bridges. Various damage detection methods have been developed to grasp the damage state of bridges. Compared with other nondestructive detection methods, the acoustic emission technology is widely used as a nondestructive detection technology, can detect dynamic defects of complex components, and can provide overall or range rapid detection for damage early warning. However, it is difficult to know the safety information of the structure in real time only by detecting the structure, and the structure needs to be monitored for health so as to know the safety information of the structure more timely.
However, the traditional acoustic emission detection method is difficult to monitor in real time. Conventional acoustic emission parametric analysis is only a limited description of the signal waveform, with which there is some deviation in the characteristics that characterize the entire acoustic emission source. The clustering analysis needs to manually make a scoring system off line and print a label value, which is not intelligent enough, and when the sample size is large, it is difficult to obtain a clustering conclusion. Emerging deep learning strategies are expected to process a large number of acoustic emission signals on line to achieve safety early warning, but the two-dimensional neural network has high computational complexity, needs special hardware for training, and is not suitable for real-time application on mobile equipment and low-power-consumption or low-memory equipment. The implementation of adaptive and compact 1D CNNs achieves higher performance than conventional deep 2D with low computational complexity. An adaptive and compact one-dimensional CNN can be effectively trained with a limited one-dimensional signal data set, rather than the massive data set required for deep 2D CNNs. Furthermore, due to the low computational load, 1D CNNs are well suited for real-time and low-cost applications, especially on smart mobile devices. The one-dimensional CNN model overcomes the problems of gradient loss and gradient explosion of a recurrent neural network, and leads the training to be quicker.
Therefore, the method and the system for monitoring the light weight and health of the small and medium bridges on line are provided by combining the 1D CNNs model with the acoustic emission signals. Compared with other acoustic emission detection technologies, the traditional acoustic emission technology for detection is used for structural health monitoring, and has the characteristics of real time and low cost.
Disclosure of Invention
The invention aims to provide a method for monitoring a medium and small bridge structure based on an acoustic emission sensor, which is used for developing a new deep learning technology and establishing a structural recognition technical process by applying the acoustic emission sensor to the monitoring of the bridge structure so as to realize light weight health monitoring and damage early warning of the bridge structure.
In order to achieve the purpose, the invention adopts the technical scheme that:
a light-weight health monitoring method and system for medium and small bridges based on acoustic emission and 1D CNNs comprises the following steps:
step one, arranging a monitoring system according to mechanical characteristics of the medium and small bridges. A health monitoring system is arranged on a bridge structure and consists of a data acquisition system and a structural health state and safety evaluation system, wherein the data acquisition subsystem is the basis of the whole monitoring system, and acoustic emission signals of the bridge structure monitored by the health monitoring system are acquired by a signal acquisition system.
A certain number of sensors are arranged at positions close to the contact surface of the pier and the bridge floor and close to the prestressed steel bars, the sensors are arranged at positions about 1 meter away from the joint position, and the sensors are installed on the bottom surface of the bridge floor. And acquiring the accompanying acoustic emission signals in the damage process of the test piece by a plurality of acoustic emission sensors arranged on the small and medium prestressed bridge. The acoustic emission signals are collected and recorded by a sensor, an amplifier and an acoustic emission analyzer, and acoustic emission characteristic parameters are extracted by the acoustic emission analyzer and input into a computer for storage.
And step two, carrying out data denoising by using Empirical Mode Decomposition (EMD). The signal processing method can decompose an input signal into the sum of a plurality of Intrinsic Mode Functions (IMFs), and the rest components meet the following two constraint conditions except that the last IMF component is a decomposition remainder: in the whole data segment, the number of extreme points and the number of zero-crossing points must be equal or the difference cannot exceed one at most; at any time, the average value of the upper envelope formed by the local maximum point and the lower envelope formed by the local minimum point is zero, that is, the upper and lower envelopes are locally symmetrical with respect to the time axis. And directly discarding a plurality of IMFs with small orders according to an energy rule, and reconstructing the rest IMFs to obtain the de-noised data.
And step three, establishing a standard classification database of the bridge damage signals by using the reduced scale model. And (3) carrying out a scale reduction test on the bridge, carrying out loading and unloading, carrying out cyclic reciprocation, measuring data by using an acoustic emission sensor, and considering the Kessel effect to enable the Kessel effect to be in accordance with the actual bridge working condition. The loading and unloading in the test were carried out in 10 groups, the last of which produced destruction. And acquiring 10 sets of acoustic emission signals generated during loading and unloading and the last set of acoustic emission signals generated during damage, and using the acoustic emission signals for subsequent model training.
In the process of breaking the bridge body of the medium and small bridge, 4 stages with the most obvious crack development change are taken, namely an uncracked stage, a crack development early stage, a crack development later stage and a breaking stage.
And determining a critical point of a crack development stage through the size change of a Kurtosis (Kurtosis) value, and realizing the division of a damage stage.
And step four, training the 1D CNNs model according to the standard classification database. Inputting a signal of a reduced scale model experiment into a one-dimensional convolution neural network model; and dividing the sampled specimen into two parts, wherein one part is used for training the model, the other part is used for testing the model, and the damage stage to which the damage belongs is judged according to the division of the 1D CNNs on the damage characteristics.
The 1D CNNs model comprises input layers, 2 convolutional layers, a maximum pooling layer and a ReLU active layer connected between the convolutional layers, a full-link layer connected behind the last convolutional layer and an output layer connected behind the full-link layer.
Full convolution and maximum pooling are adopted in the training process of the 1D CNNs model, a linear rectification function (R eLU) function is used as an activation function of a neuron, and a normalized exponential function (Softma x) function is used by a neural network output layer to calculate the probability that input data are classified into an uncracked stage, a crack development early stage, a crack development later stage and a damage stage. In the 1D CNNs model training, the neural network model parameter updating is realized by using a self-adaptive parameter optimization method, the weight of each unit is corrected by using error back propagation, and overfitt is prevented by using a dropout algorithm.
In the training stage, a sensor is used for collecting enough acoustic emission signals, one-dimensional signals obtained through EMD denoising are used as input, and a 1D CNNs model with set parameters is used for performing convolution, pooling and classification to obtain a damage identification model.
And inputting the data acquired during the scale test into the trained 1D CNNs model for damage identification to obtain a confusion matrix. The confusion matrix is used for measuring the classification accuracy of the model to the test sample and improving the reliability of the test result of the model.
And step five, applying the trained 1D CNNs model to a real bridge, and classifying according to the actually measured acoustic emission signals to perform real-time grading early warning. When the 1D CNNs model judges that the damage is in the early stage of crack development, the later stage of crack development or the damage stage, blue, orange and red early warning signals are respectively sent out.
Has the advantages that: the invention relates to a light-weight health monitoring method and a system for medium and small bridges based on acoustic emission and 1D CNNs, which can realize all-weather online monitoring on bridge structures; one-dimensional deep learning models provide a more compact and efficient solution for highly complex and diverse patterns of signal stores. The result obtained by combining the acoustic emission technology and the deep learning has the characteristics of high efficiency, low cost, high automation degree and real-time identification. The acoustic emission sensor is used for developing a signal processing technology for comprehensively monitoring the bridge structure with one or more functions and applying a deep learning model, so that the bridge structure can be comprehensively monitored and evaluated.
Drawings
Fig. 1 is a flow chart of the lightweight health monitoring method of the present invention.
FIG. 2 is a schematic view of acoustic emission sensor position and loading position in an example of the present invention.
FIG. 3 is a graph of measured acoustic emission signals, (a) measured acoustic emission signals, and (b) individual acoustic emission event signals, for an example of the present invention.
Fig. 4 is a structure of a one-dimensional Convolutional Neural Network (CNN) pattern proposed in the present invention.
FIG. 5 shows the accuracy of model identification in the present invention.
Detailed Description
The acoustic emission sensor is applied to monitoring of the bridge structure, a new deep learning identification technology is developed, a structure identification technology flow is established, and health monitoring and damage early warning of the bridge structure are achieved. The technical solution of the present invention is described in detail below by taking a certain bridge as an example, and the embodiments described herein are only used for illustrating and explaining the present invention, and are not used to limit the present invention.
As shown in fig. 1, the method and system for monitoring light-weighted health of medium and small bridges based on acoustic emission and 1D CNNs of the present invention comprises the following steps:
step one, installing a monitoring system based on an acoustic emission sensor according to the mechanical characteristics of the medium and small bridges.
Just like people need to utilize various types of large-scale medical equipment of hospitals to carry out inspection of each level on important organs in the health care process, the bridge structure also needs to carry out multi-level monitoring and identification on the important organs. The health monitoring system consists of two subsystems, namely a data acquisition system and a structural health state and safety evaluation system, wherein the data acquisition subsystem is the basis of the whole monitoring system and is used for acquiring acoustic emission signals of the bridge structure monitored by the health monitoring system through the signal acquisition system. The field data acquisition system mainly comprises a sensing system and a signal acquisition and processing system, and the structural health state and safety evaluation system is mainly composed of a deep learning model.
And 3 acoustic emission sensors are arranged at positions close to the contact surface of the pier and the bridge floor and close to the prestressed steel bars, and the sensors are respectively arranged on the bottom surfaces of the spans of the bridge floor. Fig. 2 shows a schematic diagram of the position and loading position of the acoustic emission sensor according to the embodiment of the present invention.
A plurality of acoustic emission sensors arranged on small and medium prestressed bridges are used for acquiring acoustic emission signals accompanying the damage process of the test piece. The gain is set to be 40dB, the threshold value is set to be 40dB, the sampling frequency is set to be 2.5MH z according to the field condition, and each signal comprises 2048 sampling points. And the acquired acoustic emission signals are sequentially input into a preamplifier, an acoustic emission processing module and a computer, and acoustic emission characteristic parameters are extracted by the acoustic emission processing module and input into the computer for storage.
And step two, carrying out data denoising by using empirical mode decomposition.
The signal processing method can decompose an input signal into the sum of a plurality of Intrinsic Mode Functions (IMFs), and the rest components meet the following two constraint conditions except that the last IMF component is a decomposition remainder: in the whole data segment, the number of extreme points and the number of zero-crossing points must be equal or the difference cannot exceed one at most; at any time, the average value of the upper envelope formed by the local maximum point and the lower envelope formed by the local minimum point is zero, that is, the upper and lower envelopes are locally symmetrical with respect to the time axis. And directly discarding a plurality of IMFs with small orders according to an energy rule, and reconstructing the rest IMFs to obtain the de-noised data. The waveform measured by the acoustic emission sensor is shown in fig. 3.
And step three, establishing a standard classification database of the bridge damage signals.
And (3) carrying out a scale-down test on the bridge, carrying out linear loading and unloading from 0KN, carrying out step-by-step loading by 10KN more than the last loading each time, carrying out cyclic reciprocating for 2 times, wherein each step of loading needs a period of time to ensure that cracks are fully developed, and loading until the concrete is cracked. And in the test process, data are measured by using an acoustic emission sensor, the Kessel effect is considered, and the loss stage of the medium and small bridges is divided into an uncracked stage, a crack development early stage, a crack development later stage and a damage stage according to the damage phenomenon of the medium and small bridges.
The Kessel effect is considered in the process of obtaining data by carrying out a scale test, so that the Kessel effect accords with the actual bridge working condition. The loading and unloading in the test were carried out in 10 groups, the last of which produced damage. And acquiring 10 sets of acoustic emission signals generated during loading and unloading and the last set of acoustic emission signals generated during damage, and using the acoustic emission signals for subsequent model training.
And determining a critical point of a crack development stage through the size change of a Kurtosis (Kurtosis) value, and realizing the division of a damage stage.
And step four, training the 1D CNNs model according to the standard classification database.
The 1D CNNs model comprises input layers, 2 convolutional layers, a maximum pooling layer and a ReLU active layer connected between the convolutional layers, a full-link layer connected behind the last convolutional layer and an output layer connected behind the full-link layer. The structure of a one-dimensional Convolutional Neural Network (CNN) pattern is shown in fig. 4.
In the CNN feature extraction algorithm, hierarchical layers of input data features are alternately extracted by convolutional layers and pooling layers, and a common multilayer neural network is used near an output layer. In the convolution layer, the convolution kernel performs convolution operation on the feature vector output by the previous layer, and the output feature vector is constructed by using the nonlinear activation function. The output of each layer is the convolution result of the multiple input features, and its mathematical model can be described as:
wherein M isjInputting a feature vector; l represents the number of network layers;
a convolution kernel represented as l layers;
is the network offset of layer l;
is the output of layer l;
is the input for layer l.
Maximum pooling may reduce the deviation of the estimated mean due to errors in convolutional layer parameters, the mathematical model of which may be described as:
wherein,
represents the value of t neuron in the ith feature vector of layer l; (j-1) W +1 < i < jW; w is the width of the pooling zone;
represents the value corresponding to l +1 layer neurons.
In the 1D CNNs model of this example, the size of the dimensional filter kernel is 3, the convolutional layers sequentially include 16 and 32 filters, the sub-sampling factor is 2, full convolution and maximum pooling are adopted, a linear rectification function (ReLU) function is adopted as an activation function of the neuron, and a normalized exponential function (So ftmax) function is used by the neural network output layer to calculate the probability that the input data is classified into an uncracked stage, a crack development early stage, a crack development late stage and a destruction stage. Setting a training learning rate, wherein the number of batch training samples is 50, the total iteration number is 100, the error limit is 0.1, initializing a random value of the convolution kernel weight, and initializing the bias of each layer to be 0, the weight gradient of each layer to be 0 and the bias gradient to be 0.
In the training of the 1D CNNs model of the embodiment, a training set is input into the convolutional neural network, the characteristics and errors of the convolutional neural network model are calculated layer by using an adaptive parameter optimization method, and the weights and offsets of all units are corrected by using error back propagation to extract the characteristics. Overfitt is prevented using dropout algorithm.
In the embodiment, the damage identification method based on the one-dimensional deep learning model is divided into the following steps of collecting data in a scale test according to the ratio of 7: and 3, the training stage comprises two stages of training and testing, wherein the training stage is to acquire sufficient data by using an acoustic emission microphone, take an EMD de-noised signal as input, and train by using a 1D CNNs model with set parameters to obtain a damage identification model. In the testing stage, the trained model is used for identifying the input unknown signals. And inputting the data acquired through the scale test into the trained 1D CNNs model for damage identification to obtain a confusion matrix, so that the damage identification precision is very high. The recognition result is shown in fig. 5.
And fifthly, early warning is carried out by using the obtained 1D CNNs model.
And applying the trained 1D CNNs model to a real bridge, and classifying according to the actually measured acoustic emission signals to perform real-time grading early warning. When the 1D CNNs model judges that the damage is in the early stage of crack development, the later stage of crack development and the damage stage, blue, orange and red early warning signals are respectively sent out.
Claims (6)
1. A light-weight health monitoring method and system for medium and small bridges based on acoustic emission and 1D CNNs is characterized in that: the method comprises the following steps:
step one, installing a monitoring system based on an acoustic emission sensor according to the mechanical characteristics of the medium and small bridges. A health monitoring system consisting of two subsystems, namely a data acquisition system and a structural health state and safety evaluation system, is arranged on the bridge structure.
And step two, carrying out data denoising by using Empirical Mode Decomposition (EMD). The signal processing method can decompose an input signal into the sum of a plurality of Intrinsic Mode Functions (IMFs), directly abandon a plurality of IMFs with small orders according to an energy rule, and reconstruct the rest IMFs to obtain denoised data.
And step three, establishing a reference classification database of the acoustic emission signals under the bridge damage by using the reduced scale model. And carrying out a scale-down test on the bridge, loading and unloading, measuring data by using the acoustic emission sensor, circularly reciprocating to consider the Kai ser effect, dividing damage stages of the medium and small bridges according to the damage phenomena of the medium and small bridges, and determining label information corresponding to the acoustic emission signals.
And step four, training the 1D CNNs model according to the standard classification database. Inputting a signal of a reduced scale model experiment into a one-dimensional convolution neural network model; and dividing the sampled specimen into two parts, wherein one part is used for training the model, the other part is used for testing the model, and the damage stage to which the damage belongs is judged according to the division of the 1D CNNs on the damage characteristics.
And fifthly, using the trained 1D CNNs model for early warning of the real bridge, and performing real-time safety early warning according to the classification of the actually measured acoustic emission signals. And when the 1D CNNs model judges that the damage is in the early stage of crack development, the later stage of crack development or the damage stage, sending out an early warning signal.
2. The method and system for monitoring the light-weighted health of medium and small bridges based on acoustic emission and 1D CNNs of claim 1, wherein: in the first step, a certain number of sensors are arranged at positions close to the contact surface of the pier and the bridge deck and close to the prestressed reinforcement. The acoustic emission signals are collected and recorded by a sensor, an amplifier and an acoustic emission analyzer, and acoustic emission characteristic parameters are extracted by the acoustic emission analyzer and input into a computer for storage.
3. The method and system for monitoring the light-weighted health of medium and small bridges based on acoustic emission and 1D CNNs of claim 1, wherein: and in the third step, a bridge is subjected to a scale-down test, loading and unloading are carried out, the cyclic reciprocation is carried out, acoustic emission signals are measured by using an acoustic emission sensor, and the loss stage of the medium and small bridge is divided into an uncracked stage, a crack development early stage, a crack development later stage and a damage stage according to the damage phenomenon of the medium and small bridge body. The kessel effect is taken into account in the acquisition of data with the acoustic emission sensor. I.e. load and unload were performed in 10 groups in the trial, with the last group causing damage. And acquiring 10 groups of acoustic emission signals generated by loading and unloading, and determining label information of corresponding acoustic emission signals according to the crack development stage for subsequent model training.
4. The method and system for monitoring the light-weighted health of medium and small bridges based on acoustic emission and 1D CNNs of claim 1, wherein: in the fourth step, full convolution and maximum pooling are adopted in the training process of the 1D CNNs model, a linear rectification function (ReLU) function is used as an activation function of the neuron, and the probability that input data are classified into various types is calculated by the output layer of the neural network by using a normalized exponential function (Softmax) function. In the 1D CNNs model training, the neural network model parameter updating is realized by using a self-adaptive parameter optimization method, the weight of each unit is corrected by using error back propagation, and overfitt is prevented by using a dropout algorithm.
5. The method and system for monitoring the light-weighted health of medium and small bridges based on acoustic emission and 1D CNNs of claim 1, wherein: the reduced scale model and the small bridge in the step four are structural bridges, and the measured acoustic emission signals can represent acoustic emission characteristics of a real bridge. And (4) using the acoustic emission signal sample obtained by the reduced-size model for training and testing the 1D CNNs model. And evaluating the performance of the model by adopting a confusion matrix. And acoustic emission signal samples obtained by field tests are used for verifying the 1D CNNs model.
6. The method and system for monitoring light-weighted health of medium and small bridges based on acoustic emission and 1D CNNs of claim 1, wherein: and fifthly, dividing the signals according to the trained model, and sending out early warning signals when the condition that the damage of the bridge caused by heavy vehicle passing or other accumulated damage is in the early stage of crack development, the later stage of crack development or the damage stage is monitored, or the original damage is in the early stage of crack development, the later stage of crack development or the damage stage is monitored. And when the damage is monitored to be in the early stage of crack development, the later stage of crack development and the damage stage, respectively sending out blue, orange and red early warning signals.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210172090.7A CN114660180A (en) | 2022-02-24 | 2022-02-24 | Sound emission and 1D CNNs-based light-weight health monitoring method and system for medium and small bridges |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210172090.7A CN114660180A (en) | 2022-02-24 | 2022-02-24 | Sound emission and 1D CNNs-based light-weight health monitoring method and system for medium and small bridges |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN114660180A true CN114660180A (en) | 2022-06-24 |
Family
ID=82026945
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210172090.7A Pending CN114660180A (en) | 2022-02-24 | 2022-02-24 | Sound emission and 1D CNNs-based light-weight health monitoring method and system for medium and small bridges |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114660180A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117007688A (en) * | 2023-08-07 | 2023-11-07 | 郑州大学 | Real-time filtering method for acoustic emission monitoring noise of bridge prestress steel beam damage |
| CN117647586A (en) * | 2024-01-29 | 2024-03-05 | 北京科技大学 | Dynamic monitoring and identifying method and device for coal rock damage |
-
2022
- 2022-02-24 CN CN202210172090.7A patent/CN114660180A/en active Pending
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117007688A (en) * | 2023-08-07 | 2023-11-07 | 郑州大学 | Real-time filtering method for acoustic emission monitoring noise of bridge prestress steel beam damage |
| CN117007688B (en) * | 2023-08-07 | 2024-05-28 | 郑州大学 | Real-time filtering method for acoustic emission monitoring noise of bridge prestress steel beam damage |
| CN117647586A (en) * | 2024-01-29 | 2024-03-05 | 北京科技大学 | Dynamic monitoring and identifying method and device for coal rock damage |
| CN117647586B (en) * | 2024-01-29 | 2024-04-05 | 北京科技大学 | Dynamic monitoring and identifying method and device for coal rock damage |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Eftekhar Azam et al. | Damage detection in structural systems utilizing artificial neural networks and proper orthogonal decomposition | |
| Ai et al. | Source localization on large-scale canisters for used nuclear fuel storage using optimal number of acoustic emission sensors | |
| CN110879254B (en) | Steel rail crack acoustic emission signal detection method based on improved least square generation type countermeasure network | |
| CN114660180A (en) | Sound emission and 1D CNNs-based light-weight health monitoring method and system for medium and small bridges | |
| CN114563150B (en) | Bridge health online detection module generation method, detection method, tool box and device | |
| Savino et al. | Automated classification of civil structure defects based on convolutional neural network | |
| CN114169374B (en) | Cable-stayed bridge stay cable damage identification method and electronic equipment | |
| Sattarifar et al. | Damage localization and characterization using one-dimensional convolutional neural network and a sparse network of transducers | |
| CN115953666B (en) | Substation site progress identification method based on improved Mask-RCNN | |
| Li et al. | An improved AIC onset-time picking method based on regression convolutional neural network | |
| Zhan et al. | A Novel Structural Damage Detection Method via Multisensor Spatial–Temporal Graph-Based Features and Deep Graph Convolutional Network | |
| Qin et al. | A novel physically interpretable end-to-end network for stress monitoring in laser shock peening | |
| Alfaz et al. | Bridge crack detection using dense convolutional network (densenet) | |
| CN111563455B (en) | Injury identification method based on time sequence signals and compressed convolutional neural network | |
| CN115452957B (en) | Small sample metal damage identification method based on attention prototype network | |
| CN117349693A (en) | Axle fatigue crack acoustic emission signal identification method based on PSO-DBN | |
| Joshi et al. | FCNet: Deep Learning Architecture for Classification of Fatigue and Corrosion Acoustic Emission Signal | |
| CN113607068B (en) | Method for establishing and extracting recognition model of photoacoustic measurement signal characteristics | |
| CN116070083A (en) | Steel wire rope fault intelligent recognition algorithm based on convolutional neural network | |
| Zhao et al. | A single-sensor method for structural damage localization in wind turbine blades: Laboratory assessment on a blade segment | |
| Sony | Bridge damage identification using deep learning-based Convolutional Neural Networks (CNNs) | |
| Zhu et al. | Structural damage identification considering uncertainties in nonuniform measurement conditions based on convolution neural networks | |
| Li et al. | Online Condition Assessment of Prestressed Hollow-Core Slab Bridges Based on 1d Cnn Using Acoustic Emission Technique | |
| Colace et al. | Unsupervised Learning Techniques for Vibration-Based Structural Health Monitoring Systems Driven by Data: A General Overview | |
| Hu et al. | Damage Identification and Health Assessment of Crane Girder Structure Under Moving Loads |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |